System and method for controlling motion of one or more devices

Abstract

The present disclosure provides a system and a method for controlling a motion of a device from an initial state to a target state in an environment having obstacles that form constraints on the motion of the device. The method includes executing a learned function trained with machine learning to generate a feasible or infeasible trajectory connecting the initial state of the device with the target state of the device while penalizing an extent of violation of at least some of the constraints to produce an initial trajectory. The method further includes solving a convex optimization problem subject to the constraints to produce an optimal trajectory that minimizes deviation from the initial trajectory and controlling the motion of the device according to the optimal trajectory.

We claim: 
     
         1 . A controller for controlling a motion of a device from an initial state to a target state in an environment having obstacles forming constraints on the motion of the device, the controller comprising: a processor; and a memory having instructions stored thereon that, when executed by the processor, cause the controller to:
 execute a learned function trained with machine learning to generate a feasible or infeasible trajectory connecting the initial state of the device with the target state of the device while penalizing an extent of violation of at least some of the constraints to produce an initial trajectory, wherein the learned function is a neural network trained using reinforcement learning based on a reward function that penalizes the extent of violation of the constraints while allowing to produce the infeasible trajectory;   solve a convex optimization problem subject to the constraints to produce an optimal trajectory that minimizes deviation from the initial trajectory; and   control the motion of the device according to the optimal trajectory,
 wherein the environment comprises multiple devices with a respective target state associated with each of the multiple devices, wherein the learned function is formulated for each of the multiple devices, and wherein the controller is further configured to:
 execute the learned function for each device of the multiple devices to determine an initial trajectory for each device; 
 solve a joint-optimization problem to simultaneously determine an optimal trajectory of each device, wherein the optimal trajectory of each device minimizes deviation of the optimal trajectory of the corresponding device from the corresponding initial trajectory; and 
 control motion of each device based on the corresponding optimal trajectory to achieve the corresponding target state of each device. 
 
   
     
     
         2 . The controller of  claim 1 , wherein the constraints comprise at least: a constraint of reaching the target state of the device, a device-obstacle collision avoidance constraint, an inter-device collision avoidance constraint, and a keep-in constraint. 
     
     
         3 . The controller of  claim 1 , wherein the constraints comprise at least: a static constraint with a fixed location and a dynamic constraint with a location varying in time, wherein the learned function is trained to penalize the violation of only the static constraint, and wherein the convex optimization problem is solved subject to the static constraint and the dynamic constraint. 
     
     
         4 . The controller of  claim 1 , wherein the constraints are enforced as hard constraints, based on a convex approximation of each constraint. 
     
     
         5 . The controller of  claim 1 , wherein the constraints are enforced as hard constraints, when dynamics of the device include stochastic uncertainty. 
     
     
         6 . The controller of  claim 5 , wherein the constraints are enforced as hard constraints, based on a convex chance approximation of each constraint. 
     
     
         7 . The controller of  claim 1 , wherein the device corresponds to at least one of: an autonomous vehicle, a mobile robot, an aerial vehicle, a water surface vehicle, and an underwater vehicle. 
     
     
         8 . A method for controlling a motion of a device from an initial state to a target state in an environment having obstacles forming constraints on the motion of the device, the method comprising:
 executing a learned function trained with machine learning to generate a feasible or infeasible trajectory connecting the initial state of the device with the target state of the device while penalizing an extent of violation of at least some of the constraints to produce an initial trajectory, wherein the learned function is a neural network trained using reinforcement learning based on a reward function that penalizes the extent of violation of the constraints while allowing to produce the infeasible trajectory;   solving a convex optimization problem subject to the constraints to produce an optimal trajectory that minimizes deviation from the initial trajectory; and   controlling the motion of the device according to the optimal trajectory,   wherein the environment comprises multiple devices with a respective target state, wherein the learned function is formulated for each of the multiple devices, and wherein the method further comprises:
 executing the learned function of each device to determine an initial trajectory for each device; 
 solving a convex optimization problem of each device to determine an optimal trajectory for each device, wherein the optimal trajectory of each device minimizes deviation of the optimal trajectory from the corresponding initial trajectory; and 
 controlling motion of each device based on the corresponding optimal trajectory to achieve the corresponding target state of each device. 
   
     
     
         9 . The method of  claim 8 , wherein the constraints comprise at least: a constraint of reaching the target state of the device, a device-obstacle collision avoidance constraint, an inter-device collision avoidance constraint, and a keep-in constraint. 
     
     
         10 . The method of  claim 8 , wherein the constraints comprise at least: a static constraint with a fixed location and a dynamic constraint with a location varying in time, wherein the learned function is trained to penalize the violation of only the static constraint, and wherein the convex optimization problem is solved subject to the static constraint and the dynamic constraint. 
     
     
         11 . The method of  claim 8 , wherein the device corresponds to at least one of: an autonomous vehicle, a mobile robot, an aerial vehicle, a water surface vehicle, and an underwater vehicle. 
     
     
         12 . A non-transitory computer-readable storage medium embodied thereon a program executable by a processor for performing a method for controlling a motion of a device from an initial state to a target state in an environment having obstacles forming constraints on the motion of the device, the method comprising:
 executing a learned function trained with machine learning to generate a feasible or infeasible trajectory connecting the initial state of the device with the target state of the device while penalizing an extent of violation of at least some of the constraints to produce an initial trajectory, wherein the learned function is a neural network trained using reinforcement learning based on a reward function that penalizes the extent of violation of the constraints while allowing to produce the infeasible trajectory;   solving a convex optimization problem subject to the constraints to produce an optimal trajectory that minimizes deviation from the initial trajectory; and   controlling the motion of the device according to the optimal trajectory,   wherein the environment comprises multiple devices with a respective target state associated with each of the multiple devices, wherein the learned function is formulated for each of the multiple devices, and wherein the method further comprises:
 executing the learned function for each device of the multiple devices to determine an initial trajectory for each device; 
 solving a joint-optimization problem to simultaneously determine an optimal trajectory of each device, wherein the optimal trajectory of each device minimizes deviation of the optimal trajectory of the corresponding device from the corresponding initial trajectory; and 
 controlling motion of each device based on the corresponding optimal trajectory to achieve the corresponding target state of each device.